Reimagining Success and Failure: Equitable Assessment Practices in an Age of Artificial Intelligence

Synthesis: Dollinger and Nieminen (2026) present a conceptual paper arguing that GenAI's disruption of Assessment is an opportunity — not a threat — to fundamentally reimagine how student success and failure are defined in higher education. Rejecting the containment approaches (surveillance, invigilation, AI Detection) that dominate current institutional responses, they contend that GenAI exposes what was already true: that Assessment practices have long been broken, sorting and ranking students in ways that systematically disadvantage those from equity-deserving backgrounds. The paper proposes reimagining success along two dimensions — a shift from individualistic to distributed understandings of knowledge, and a move from predetermined standards toward agentic assessment that positions students as active participants in defining success — with profound implications for Authentic Assessment and grading.

Key Findings

• Assessment was already broken; GenAI just made it undeniable: Decades of Assessment research showing inequity remain unaddressed or even exacerbated. Universities have long relied on assessments that sort, rank, and grade students, measuring reproduction of knowledge under artificial constraints rather than how knowledge is actually created or used. GenAI challenges individual authorship so fundamentally that maintaining current practices is no longer possible without explicit justification.
• Rejecting both dominant equity approaches: The paper critiques (1) the accommodations model, which places the burden to adapt on individual "at risk" students, and (2) structural critiques that reframe assessment as disabling students but still leave intact the zero-sum logic that "for one student to succeed, someone else needs to fail" — what Nieminen (2024) calls the paradox of inclusive assessment.
• From individualistic to distributed success/failure: The myth that valid knowledge resides within a single individual privileges students who excel in isolated, high-pressure performance (e.g., closed-book, time-constrained exams, which show differential outcomes by gender, socio-economic status, race, and disability) while disadvantaging collaborative, resource-rich learners. GenAI makes the inherently social, distributed nature of knowing impossible to ignore, reframing group work and collaborative process documentation as evidence of "integrated knowledge networks."
• From predetermined to agentic success/failure: Standardization rooted in industrial-era values fixes learning outcomes before students arrive, positioning them as performers rather than agents. Agentic assessment structurally enables students to participate in defining what counts as success and failure — negotiating formats, criteria, and timelines — and validates diverse ways of knowing within academic structures.
• Reframing around detection of learning, not detection of cheating: Rather than asking "what can't GenAI do," educators should ask what students should learn through assessment: navigational choices, reflections, and judgments (i.e., "detecting learning rather than detecting cheating," per Ellis & Lodge 2024).
• Grading reform (the "elephant in the room"): Grades are the most powerful cultural marker of success/failure. The paper advocates competency-based or pass/fail frameworks (which can be as rigorous as rubrics), process-documented portfolios, "persuasive portfolios," navigational-capability scoring, and student self-grading — noting that we might "grade our grading systems rather than our students."

What this means for practice

• Instructors. Ask what students should learn through the task instead of what GenAI cannot do — navigational choices, reflection, and judgment, or "detecting learning rather than detecting cheating" — and let that question drive the design.
• Instructors. Drop detection-based defenses such as AI-detection tools and invigilated high-stakes exams: the paper cites their methodological limitations, false positives, and violations of procedural fairness, and argues that containment preserves the sorting logic that made Assessment inequitable in the first place.
• Designers. Make success distributed rather than individual: assess group work by the documented synthesis across perspectives — an integrated knowledge network — rather than by a single isolated performance under artificial constraints, which is the format that privileged some students and disadvantaged collaborative, resource-rich learners.
• Designers. Make assessment agentic by letting students negotiate formats, criteria, and timelines, and consider the grading alternatives the paper sets out — competency-based or pass/fail frameworks, process-documented portfolios, and student self-grading — with the aim of grading the grading system rather than the students.
• Administrators. Begin with localized experimentation inside individual courses, because accreditation requirements, grading policies, and external ranking systems are real constraints, and treat course-level reform as the evidence-generating entry point the authors call for.

Limitations

• This is a conceptual paper: it reports no sample, intervention, or outcome data, so it can argue what equitable assessment should look like but cannot show that distributed or agentic designs improve equity or learning.
• The authors acknowledge that their discussion addresses how GenAI might support more inclusive assessment with limited attention to broader equity concerns of AI systems — bias, potential labor displacement, and the concentration of technological power in private entities.
• They flag the risk that reform disadvantages students who benefit from clear structure and predetermined expectations, and that students may feel uncertain without numerical feedback and need scaffolded support to build intrinsic motivation; both are left unresolved in the paper's own account.
• Grading reform lies largely outside the authors' control: institutional grading policy, accreditation, and external ranking systems constrain implementation, and no evidence is offered yet on how the proposed changes affect equity outcomes, which the paper names as a target for future empirical research.

Citation

Dollinger, M., & Nieminen, J. H. (2026). Reimagining Success and Failure: Equitable Assessment Practices in an Age of Artificial Intelligence. Journal of University Teaching and Learning Practice, 23(1). (CC BY-ND 4.0.)